WildActor: Unconstrained Identity-Preserving Video Generation
This addresses the need for production-ready human video generation with consistent identity, though it appears incremental as it builds on prior video generation methods.
The paper tackles the problem of generating human videos where digital actors maintain consistent full-body identities across dynamic shots, viewpoints, and motions, and presents WildActor, a framework that achieves this by leveraging a new large-scale dataset and novel attention and sampling mechanisms, surpassing existing methods in challenging settings.
Production-ready human video generation requires digital actors to maintain strictly consistent full-body identities across dynamic shots, viewpoints and motions, a setting that remains challenging for existing methods. Prior methods often suffer from face-centric behavior that neglects body-level consistency, or produce copy-paste artifacts where subjects appear rigid due to pose locking. We present Actor-18M, a large-scale human video dataset designed to capture identity consistency under unconstrained viewpoints and environments. Actor-18M comprises 1.6M videos with 18M corresponding human images, covering both arbitrary views and canonical three-view representations. Leveraging Actor-18M, we propose WildActor, a framework for any-view conditioned human video generation. We introduce an Asymmetric Identity-Preserving Attention mechanism coupled with a Viewpoint-Adaptive Monte Carlo Sampling strategy that iteratively re-weights reference conditions by marginal utility for balanced manifold coverage. Evaluated on the proposed Actor-Bench, WildActor consistently preserves body identity under diverse shot compositions, large viewpoint transitions, and substantial motions, surpassing existing methods in these challenging settings.